VBSF001 / README.md
thotnd's picture
Update README.md
6ad3bf6
metadata
license: mit

Vietnamese Text-To-Speech dataset (VBSF001-v1.0)

The audio is crawled from audiobooks on YouTube. The audio is NOT for commercial use.

The text is labeled by VinBrain JSC. The text is in the public domain.

Dataset size: 1.28GB.

Total audio duration: 9.5 hours.

Text-audio samples

  • Sample 1:
    • Audio: file1
    • Text: song cũng từ buổi tối bất hạnh ấy, một nỗi buồn ghê gớm xâm chiếm nhà vua và thường xuyên lộ ra nét mặt.
  • Sample 2:
    • Audio: file2
    • Text: theo luật pháp nước nhà, sa-dơ-năng không được chia quyền, đành phải sống như một người dân thường.

Download

Get the dataset from here: link.

VBSF001 directory structure:

VBSF001
├── metadata.csv
└── wavs
    ├── 1.wav
    ├── 2.wav
    ├── 3.wav
    ...

Statistics

  • Total Clips: 6,835
  • Total Words: 121,423
  • Total Characters: 528,088
  • Total Duration: 9:32:29
  • Mean Clip Duration: 5.03 sec
  • Min Clip Duration: 0.58 sec
  • Max Clip Duration: 14.09 sec
  • Mean Words per Clip: 17.76
  • Distinct Words: 3,527